EOMA68 Computing Devices

An Earth-friendly way to easily upgrade and fix your own computer

Apr 14, 2018

Project update 61 of 75

DDR3 RAM and a Libre RISC-V SoC

by Luke Kenneth Casson Leighton

There really wasn’t much to say until today, when finally, after literally months of enquiries, an obscure manufacturer of 4 gigabit 78FBGA 1.5v DDR3 x8 RAM ICs with actual stock and a reasonably sane price of $3.30 USD was found. I will not say who they are yet just in case someone else wants them, demand is that high and prices that unstable. They only have 2,000 in stock, meaning that only 500 boards (four RAM ICs per board) could be made up, but that’s OK: a continued search for another 2,000 RAM ICs can go ahead in parallel. This manufacturer may even make another run in the meantime: they have however made it clear that their primary focus is the 96FBGA DDR3 x16 RAM ICs.

So what the heck is going on? What’s the damn fuss, and why the hell haven’t the boards been manufactured already when everything is available? Surely it’s possible to get hold of DDR3 RAM ICs, right? Well… yes and no. I suspect that what’s happened recently is that Apple and probably the introduction of other products has driven pricing down on DDR4 to the point where, pretty much across the board, every major manufacturer of DDR3 RAM ICs has stopped making them. Simultaneously. This places companies like Freescale, with the iMX6 (which is supposed to have a 19-year lifetime), and Texas Instruments, with their OMAP and other series SoCs that only have DDR3 interfaces, in severe jeopardy.

What’s pricing like in Shenzhen markets at the moment for 4 gigabit 1.5v DDR3 x8 RAM ICs? Hynix RAM - at the last quote I received - was four dollars fifty cents. 27 RMB. With no guarantee of supply!. Why does that matter? because as previously mentioned in every one of the past few updates, there’s absolutely no point buying the RAM ICs until samples are tested, and with no guarantee of supply there’s no point in going ahead with the testing.

So, yes, we could have possibly gone ahead and spent the entire remaining budget on RAM ICs that were not even guaranteed to work ($18,000 USD on RAM ICs alone). I would prefer, putting it in a spectacularly understated fashion, to only spend $13,200 USD on RAM ICs and do so only when I know there is a 100% proven chance that the purchased RAM ICs would actually do the job, i.e., actually, like… work.

It’s one thing after another, isn’t it? Oh, and now we have the idiotic President of the United States thinking that he can get into a pissing contest with China over trade tariffs. China responded instantly by raising by up to 20% the import duties on a weird list of products (they included "pork sausages" I believe, amongst other oddities) that makes clear that whoever is in charge of China’s Customs and Excise has a hot-team on standby monitoring precisely which products from which countries are most popular, thus permitting them, within days, to deep-six efforts by any country that tries a trade war. The complaints from USA agriculture associations have already been made public, so there is a small chance that by the time the boards are manufactured it will not be necessary to completely rethink the distribution strategy.

The Libre RISC-V SoC

I’ll be brief on this one (despite appearances: a lot has been going on): there’s a new website with associated git repository and bugtracker, and I’m now extremely busy coordinating various efforts to get peripherals tracked down (or designed). I’m also on the RISC-V mailing lists coming to grips with the instruction set and with the way that the (proprietary, closed) RISC-V Foundation operates. It’s not good news, I’m afraid, despite the fact that they’ve achieved such an enormous amount, and inspired so many people to work together.

There are signs however that people are listening, and my… communication style and learning difficulties seem to be tolerated. Some people have told me that they find the change of style to be informative, refreshing and amusing: others have been angered by my blunt honesty at informing them of the consequences of how the RISC-V Foundation is run: there are clear systemic failures which need to be addressed, otherwise there will be a hard-fork of the entire RISC-V ecosystem.

What’s particularly challenging is that I am neither permitted to report bugs (because they use GitHub) nor join the RISC-V Foundation due to clear cognitive dissonance and ethical conflict of interest in Sections 1.9 and Section 5 of the RISC-V Member Charter Agreement. Not only that, but other people on the "outside" - those who have spent considerable time helping to make RISC-V better - have been shut out due, for example, to a corporation pulling out of the RISC-V Foundation, their employee being the chair of one of the working groups, and the entire working group was shut down. The control exerted by corporate interests has therefore not gone unnoticed.

The good news is that there is an anonymous sponsor willing to help cut through this. The anonymous sponsor has - believe it or not - paid for an SDCard Association membership so that, under NDA, a contractor can be paid to implement a BSD-licensed SD/MMC and eMMC hard macro! Various other plans are underway, I can’t say more at the moment.

If there are any people who have experience with MyHDL, Verilog, Bluespec (BSV), cocotb, FPGA prototyping, processor architecture and design, please do get in touch: there’s a budget available for the right people willing to release or work on BSD-licensed RTL, to help get a truly libre-licensed RISC-V Soc with 3D graphics and video decode capability into production. Also I will be in Barcelona for the RISC-V Workshop, 7th to 10th May, if anyone happens to be nearby (or attending). I am inviting people to an informal BoF to discuss abstraction of parallelism, as there is functional duplication in both the vector extension, and the SIMD extension, and also companies such as VectorBlox have developed their own custom parallel engine.

Almost lastly: many many apologies to the Crowd Supply team, I really wanted to come and meet you at Teardown 2018, which is on the 11th to 13th. You’ve been so supportive and helpful for so long. Lastly, thank you to everyone who has been sponsoring me through Liberapay, it is really appreciated. The weekly donations of $0.01 are particularly funny, and are appreciated even more, given that, collectively they add up over time. If a thousand people donate \$0.01 a week that’s \$10 a week, which is not insignificant, so thank you.

Sign up to receive future updates for EOMA68 Computing Devices.

Subscribe to the Crowd Supply newsletter, highlighting the latest creators and projects